| { | |
| "model_name": "Qwen3-5M-GPT2", | |
| "model_type": "Qwen3ForCausalLM", | |
| "tokenizer": "gpt2", | |
| "dtype": "bfloat16", | |
| "vocab_size": 50257, | |
| "hidden_size": 384, | |
| "num_layers": 12, | |
| "num_attention_heads": 6, | |
| "num_key_value_heads": 2, | |
| "head_dim": 64, | |
| "intermediate_size": 1024, | |
| "max_position_embeddings": 8192, | |
| "rope_theta": 500000, | |
| "sliding_window": 512, | |
| "use_sliding_window": true, | |
| "layer_types": [ | |
| "sliding_attention", | |
| "sliding_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "sliding_attention", | |
| "sliding_attention", | |
| "full_attention", | |
| "sliding_attention", | |
| "sliding_attention", | |
| "sliding_attention", | |
| "full_attention" | |
| ], | |
| "max_window_layers": 9, | |
| "query_pre_attn_scalar": 256, | |
| "parameters": 38184192, | |
| "tie_word_embeddings": true, | |
| "positional_encoding": "rope", | |
| "normalization": "rmsnorm", | |
| "activation": "swiglu", | |
| "xsa_enabled": false, | |
| "xsa_paper": "arxiv 2603.09078" | |
| } |